The Semiconductor Goldrush - India’s bid

#Semiconductor #Conductors #Chips
The Semiconductor Goldrush - India’s bid

Not too long ago, our news feed and the general public psyche were filled with news of the chip shortage. The news brought limelight to something so essential that saying ‘the whole world runs on it’ would hardly be an overstatement: Semiconductors. The world faced a severe semiconductor shortage in 2021 and 2022, hampered the production of cars and everything else that uses semiconductor inputs. In this two-part series, let’s briefly look at how the industry works, why it is so difficult to tap into and what India has done to tap into this market.

Semiconductors: Impossible to create, Trivial to duplicate 

Interesting title, right? It’s not just clickbait; by the end of this segment, you’ll resonate with it too. What will you call a sector in which it is not uncommon for industry revenues to double over the course of a couple of years just to fall by 50 per cent over the few years after that: Volatile. 

This volatility is caused by the unique characteristics of the semiconductor industry, which make this dramatic boom-bust cycle all but inevitable. And we will almost certainly see a crash over the next few years. 

Technically a semiconductor is made up of silicon that conducts electricity to a greater extent than an insulator such as glass but to a lesser extent than a conductor such as copper. Semiconductors are vital components in computer chips, smartphones, cars or any other digital computer device. On the surface, this seems like it should be a great industry as technology becomes increasingly advanced every year, and the number of chips required in each device and, thus, the total demand for chips increases. However, that is far from the truth; global semiconductor revenues are highly volatile, with year-to-year swings of more than 20 per cent, either up or down. Semiconductors such as DRAMs allow computers to access data in their memory storage and are widely used in personal computers, smartphones and video game consoles. Unsurprisingly, it is very complicated to manufacture, and there are tremendous economies of scale involved. Hence, the market is dominated by three companies in America's micron technologies and two South Korean companies, SK Hynix and Samsung semiconductors, the dram industry specifically (A type of semiconductor). 

Revenues are extremely volatile; For example, from 2016 to 2018, quarterly revenue tripled from about 9 billion dollars to 27 billion dollars the following year, and revenue tanked from almost 45 to 15 billion these dramatic swings are hard to make sense of at first since the end products that use dram inputs experience slow steady growth during these periods so dram volumes shipped did not face any significant volatility. It turns out that almost all of the decrease in revenue was driven by pricing when prices were down. Even if you ship the same number of units as in 2016, your revenue will also fall by 40 this extreme volatility in DRAM pricing is nothing new to the industry. In the early 2000s, the integrated circuit engineering corporation published a report examining the boom-bust cycles of a dram from 1977 through 1996 the dram market saw four boom bus cycles which saw industry revenues increasing between four and six-fold before tanking 25 to 50 per cent. However, during this same time, DRAM units shift actually increased every single year these increases were driven by strong demand growth from personal computers, which use dram inputs the boom and bust cycles were driven almost entirely by supply, not demand, when supply was relatively low the average selling price would skyrocket, leading to higher revenues when supply increased, the industry would ship more units, but the average selling price would decline dramatically, causing revenues to the tank.

The boom and bust cycle happens in the following few steps:

  1. Steadily Increasing demand for semiconductors causes prices to rise 
  2. Manufacturers build new capacity to take advantage of the high prices 
  3. This new capacity causes supply to increase, causing prices to tank
  4. Manufacturers to stop building new capacity as it is no longer profitable at this low prices demand eventually catches up to capacity,
  5. And the whole cycle repeats itself 

Now the question arises, Why is it the case that semiconductor producers over-build capacity when prices are high?

The boom-bust cycle is bad for these companies as they often lose huge amounts of money during the bust phase. And if not, as if these companies are not highly sophisticated or don’t understand the boom-bust cycle better than anyone else. 

There are two main reasons for overbuilding, even after all this common knowledge:

  1. Semiconductor factories have tremendous economies of scale. They require expensive machinery that must be utilised constantly to cover the sunk costs (Cost of the expensive machines and equipment) a small factory will be inefficient to take in, as the fixed cost will be absorbed by only a small number of units produced.
  • The only way to produce dram cheaply is to build giant factories that cost billions of dollars when prices are high. Micron, Samsung and all others built these massive new factories to take advantage of this. It allows them to make billions of dollars of profits for a few years, but when all the capacity comes online, the supply becomes too great and prices tank.

  1. The marginal cost of producing the chips is very low; once the factory runs, the cost of producing 50 or 100 chips is almost the same. So, even if it is unprofitable at low prices, they still run close to full capacity to cover as many fixed costs as possible. Since the whole industry is producing too much at this point, if any individual company reduces its capacity, it will only have a small impact on prices. In an imaginary world, if it would be possible for Micron XK Hynix and Samsung to meet and agree to reduce their output when prices were too low, they could raise prices which would increase revenue and profitability for the whole industry. (Tch Tch OPEC and Oil). Well, OPEC is allowed to do this because they are sovereign countries making deals with each other. Thus, they do not have to abide by anti-collusion laws. However, if semiconductor companies tried to do a similar deal to limit capacity, this would be considered illegal collusion.

The economics of the semiconductor industry make boom-bust cycles inevitable there's not really anything the companies can do to avoid it, which brought the industry to its knees when the pandemic struck in 2020.

India’s tryst with semiconductors: Past 

India is a semiconductor design powerhouse. Nearly every major semiconductor company has a presence in India, designing some of the most advanced chips in the world. But once those designs are completed, they are sent to the United States, China, South Korea or Taiwan to fabricate.  It begs the question. Why can't India fabricate any of the chips they lay out? 

In the 1980s, the Indian government attempted to follow China, Taiwan, Malaysia, Korea, and Singapore in creating their own semiconductor manufacturing national champion or Semiconductor Laboratory or SCL. The national champion that failed. 

Microprocessors and other semiconductors as the potential foundation for a new revolution; SCL's goal was to eventually design and manufacture leading-edge conductors headquartered in the planned city of Mohali in Punjab. But the area was up and coming; it was emerging as an electronics industry centre, with major units like Punjab Wireless Systems and Punjab Communications Limited based there.  To fill the ranks, SCL hired young graduates from top technical universities like the Indian Institutes of Technology and the Indian Institute of Science in Bangalore. They were also able to recruit from esteemed Indian electronics companies like Bharat Electronics Limited, the Indian state-owned aerospace and defence electronics company. 

There are a lot of factors that go into creating a successful, high-performing semiconductor factory, critical factors being, Financial capital, Human capital, Government/infrastructure, and Manufacturing technology. In India, all of the machinery required was imported from places like Europe or Japan. Import duties and documents have to be addressed and streamlined. SCL wanted to be a leading semiconductor manufacturer. And as I mentioned, the SCL Climbs the Ladder playbook for doing so is pretty well-established. They had some financial capital, smart people, and government buy-in. Now was acquiring the older technology. 

You strike a technology transfer deal with a more advanced company. Get them to teach you how to do it. A world-class semiconductor companies know this, of course. And so they very closely guard their technology process nodes and trade secrets. Technology transfer agreements are almost always for older processes and never for the leading-edge stuff. At SCL's founding in 1984, they could license a 5-micron process technology from American Microsystems Inc. Shortly thereafter, SCL was able to acquire process technology from two other companies as part of a deal to manufacture electronics components for them. The first deal was with the American industrial automation company Rockwell for the purpose of making their 2560G microprocessor. The second deal was with the Japanese firm Hitachi for the purpose of producing components for their electronic wristwatch. The company also performed third-party assembly duties for electronics brands. They assembled the BBC Acorn computer for the Indian Government's Computer Literacy and Studies in Schools program. These technology transfer agreements allowed SCL personnel to travel to the US and Japan for in-person training and learning. SCL quickly disseminated this info to the rest of the company. They augmented this trade knowledge with university academic partnerships. Using these resources, the company advanced very quickly from the 5-micron process technology down to a 0.8-micron process in the late 1980s (The lower the number, the more advance it is). 0.8 microns, or 800 nanometers, was first achieved in 1987 by leading companies like NTT, Toshiba, and Intel. So at this point, SCL was one semiconductor generation behind the leading edge. It seemed possible that India could achieve its goal of being a global semiconductor manufacturer within a decade. 

Those hopes ended in 1989 when a devastating fire broke out at SCL. The cause of the fire Lost Progress remains unknown. A few sources have claimed, without evidence, that it was arson. The fire was a devastating setback for India's semiconductor manufacturing efforts. Fires are devastating because the burning chemicals are toxic and release corrosive gases. The fire then triggers the water sprinkler system, causing even more damage. It would take until 1997-8 years-and substantial financial investment north of $50 million - before production finally restarted. By then, new entrants like TSMC (founded in 1987) and Samsung had entered the race. And they quickly raced ahead of the rest of the world, capturing critical global market share and scale. India lost untold amounts of progress. The government subsequently lurched through failed proposal after failed proposal in order to try to make up for lost ground. They wanted to sell the fab. But potential private investors could not agree with the Indian government on terms. They then retooled the fab from making chips for telephone exchanges to making chips for smart cards. This didn’t really take off, either. Then finally, in 2005, the company was restructured as an R&D centre within the Department of Space. SCL would be renamed to mean “Semiconductor Lab”. This finally ended SCL's chances of being a competitive commercial entity. But the company had long been out of the race.  Much of its revenue came from government contracts. SCL's offerings were not even competitive in the domestic Indian market, let alone abroad. But even with the government as a captive customer, the company could not turn an attractive profit. 

There were no economies of scale. In the 2005-2006 financial period, the company produced 1,000 6-inch wafers but had 20x the installed capacity. In 1999-2000, SCL made $14 million in revenue and just $400,000 in profit. In 2005-2006, its final year as a company, SCL made $3.5 million in revenue and turned a loss of $2 million USD. The year before, the company lost a stunning $5.6 million USD. Its restructuring was a kind mercy. Today, SCL mostly does R&D work with its old 6-inch wafer fab. Recently in 2019, they announced that they are able to accept chip designs at the 180 nm node. This is nowhere near the leading edge. But, of course, there are plenty of viable commercial cases for fabbing semiconductors at higher nodes. The more critical issue has to do with extremely sluggish development. It took SCL nearly a decade to reach that 180nm node. They have been working on it since at least 2011 when they paid millions to Israel's TowerJazz Semiconductor for a fabrication unit. SCL in the early 1980s had the advantage in that everyone else in the rest of the world, The Indian Challenge Today, was not too far ahead of them. Taiwan and China had yet to move into semiconductors. And the equipment back then was a lot cheaper. But even before the fire in 1989, SCL was starting to experience the financial strain of competing in the semiconductor business. The costs of going to 800 nanometers forced them to drop the BBC Acorn Computer project. Talented SCL personnel were constantly leaving for better jobs in the private sector or abroad. For India, the sheer amount of capital needed to build competitive, productive fabs has been cited as the single most significant obstacle in establishing a viable semiconductor manufacturing industry. 

Today's leading-edge chip factories regularly run up into the tens of billions of dollars. This trend really started to ramp up at the 14nm node generation, and it is only getting worse. Even for the biggest private Indian entities - companies like Tata, Reliance and such - this is hard to swallow. Reliance Jio initially spent $15 billion USD on their nationwide LTE data network over the span of four years. TSMC will spend $20 billion on just one Gigafab over two years. And right now, they are building three. All that investment has to pay for itself. And in the volatile electronics world, that is not always a given. Regardless, the Indian government has since attempted to revive the country's Attempts at Revival semiconductor manufacturing efforts. All of these efforts have failed. 

For instance 2006, the country announced a $3 billion "Fab City" project for manufacturing. AMD had been interested in putting an assembly and test facility there until bad industry conditions shut it down. In 2013, the Indian government lifted customs duties on all imports of parts and machinery related to semiconductor manufacturing. But this did not seem to have helped to jumpstart any foundry efforts. In 2014, India approved a proposal from two investor groups to build fabs in India. Both projects together would cost about $10 billion. The government would have provided ample financial support - up to 25% of the total cost in interest-free loans, tax breaks and subsidies. The first consortium had some notable names - Jaiprakash or JP Associates Ltd, Israel's TowerJazz, and IBM. JP Associates is a large Indian conglomerate in the construction, power, and real estate businesses. But in 2016, JP and its group pulled out of the project. The firm had a lot of debt and said that a semiconductor plant was not commercially viable. The second consortia was led by a company called Hindustan Semiconductor Manufacturing Corporation or HSMC.

The investor consortium had tapped chipmaking expertise from European chipmaker STMicroelectronics and Malaysian state operator Silterra. This second team also did not pan out. In 2019, the government cancelled HSMC's 2-year-old letter of intent. The consortium had to submit documents to demonstrate commitment but failed. An interesting coincidence. A year later, another semiconductor company named HSMC, this time standing for Hongxin Semiconductor Manufacturing Company, raised millions to start a fab in the Chinese city of Wuhan. A similar lack of execution - in addition to a pandemic - ended up trashing that too. Since then, no serious Indian proposal has emerged. And that is despite the current market conditions. I want to re-emphasise that India retains world-class chip design capabilities. Tens of thousands of Indian engineers work directly in VLSI design, and their chips are taped out in leading-edge fabs around the world. Virtually all the world's biggest fabless chip companies have an Indian presence. Since the mid-1980s, the design and manufacture of chips have split apart. It is no longer necessary to do both at the same time. And increasingly, few companies and economies are capable of doing so anyway. So India's powerful advantages in design should be commended. SCL's failure to stand on its own as a commercial entity has had long-term consequences. Today, India lacks substantial semiconductor fabrication capacity. There is hardly any semiconductor fabrication capacity at all. 

As of this writing, 100% of India’s chips - logic, memory, all of it, have to be imported from abroad
. In 2019, India imported $21 billion worth of semiconductors, according to the India $21 billion India Semiconductor Imports Electronics and Semiconductor Association. This number is growing at about 15% a year. In 2020, the country fell prey to the worldwide semiconductor shortage, disrupting various critical Indian businesses. For instance, India is the world's second-largest smartphone maker. That lucrative industry grinds to a halt when the right chips are unavailable. Furthermore, a significant portion of these imports - 37% in 2019 or $7 billion worth-came from China. This has critical geopolitical repercussions Sino-Indian relations today are tense, and further disruptions may occur. 

There is no shortage of ideas and human talent or possible approaches to bring semiconductor manufacturing back to Indian shores. But there must be consistent, long-term direction from the very top and fountains of money committed to that direction. For India, the market has long passed the phase when a private entity can seed a thriving semiconductor manufacturing industry without outside help. The government has to do more to catch up. We’ll see how that plan is faring in the next article!

Share in Your Network

Join Our Founders Community

Meaningful discussions, ask questions, and share experiences with fellow founders

Be a Part of Top 1%

Similar Blogs